From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 14339 invoked by alias); 29 Jul 2006 18:54:40 -0000 Received: (qmail 14326 invoked by uid 22791); 29 Jul 2006 18:54:38 -0000 X-Spam-Check-By: sourceware.org Received: from ip-160-218-156-177.eurotel.cz (HELO host0.dyn.jankratochvil.net) (160.218.156.177) by sourceware.org (qpsmtpd/0.31) with ESMTP; Sat, 29 Jul 2006 18:54:29 +0000 Received: from host0.dyn.jankratochvil.net (localhost [127.0.0.1]) by host0.dyn.jankratochvil.net (8.13.7/8.13.7) with ESMTP id k6TIrYRj017857; Sat, 29 Jul 2006 20:53:34 +0200 Received: (from lace@localhost) by host0.dyn.jankratochvil.net (8.13.7/8.13.7/Submit) id k6TIrHEt017856; Sat, 29 Jul 2006 20:53:17 +0200 Date: Sat, 29 Jul 2006 18:54:00 -0000 From: Jan Kratochvil To: Daniel Jacobowitz Cc: gdb-patches@sourceware.org Subject: [patch] Linux MAY_FOLLOW_EXEC #2 [Re: RFC: Fix crash on i386 (%gs-)threaded programs using execve(2)] Message-ID: <20060729185317.GA16200@host0.dyn.jankratochvil.net> References: <20060614105510.GA12067@host0.dyn.jankratochvil.net> <20060614142552.GA15021@nevyn.them.org> <20060615203519.GA9603@host0.dyn.jankratochvil.net> <20060721181556.GA9150@lace.redhat.com> <20060721184421.GA22820@nevyn.them.org> <20060722123102.GA1936@lace.redhat.com> <20060724190332.GA13612@nevyn.them.org> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="3MwIy2ne0vdjdPXF" Content-Disposition: inline In-Reply-To: <20060724190332.GA13612@nevyn.them.org> User-Agent: Mutt/1.4.2.1i Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org X-SW-Source: 2006-07/txt/msg00421.txt.bz2 --3MwIy2ne0vdjdPXF Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-length: 2358 Hi Daniel, On Mon, 24 Jul 2006 21:03:32 +0200, Daniel Jacobowitz wrote: ... > The reason I find it so disorienting is this: > > % gdb file1 > > (gdb) run > [starts file1] > [file1 execs file2] > [file2 exits] > > (gdb) run > [file2 starts instead of file1!] > > I don't know if it should do that or not. I tend to use "run" a lot > and want to get back to the beginning of my debug session. Included restoration patch using make_run_cleanup(), session log below. That waitpid(3) WNOHANG smells there for me but it does not hurt the testsuite. Should I investigate it more? 2006-07-29 Jan Kratochvil * inf-ptrace.c (inf_ptrace_mourn_inferior): waitpid(3) with WNOHANG to avoid lockup on the introduced TARGET_WAITKIND_EXECD handling. * infrun.c (follow_exec): Unconditionally enabled by MAY_FOLLOW_EXEC. Provide restoration of exec_bfd and symfile_objfile for any new "run". * linux-thread-db.c (thread_db_wait): Handle TARGET_WAITKIND_EXECD. * linux-thread-db.c (thread_db_mourn_inferior): Turn off threading. * foll-exec.exp: Uncoditionally enabled for all platforms. Relaxed regex to apply besides HP-UX also for GNU/Linux backtrace. Thanks, Jan ------------------------------------------------------------------------------ (gdb) file execve Reading symbols from /tmp/execve...done. Using host libthread_db library "/lib/libthread_db.so.1". (gdb) catch exec Catchpoint 1 (exec) (gdb) set args gdb-entered parameters (gdb) run Starting program: /tmp/execve gdb-entered parameters Executing new program: /tmp/hello [Switching to process 16524] Catchpoint 1 (exec'd /tmp/hello), 0x44031840 in ?? () (gdb) show args Argument list to give program being debugged when it is started is "gdb-entered parameters". (gdb) info files Symbols from "/tmp/hello". Unix child process: Using the running image of child process 16524. While running this, GDB does not access memory from... Local exec file: `/tmp/hello', file type elf32-i386. [...] (gdb) c Continuing. [New process 16524] hello world Program exited normally. (gdb) info files Symbols from "/tmp/hello". Local exec file: `/tmp/hello', file type elf32-i386. [...] (gdb) run Starting program: /tmp/execve gdb-entered parameters Executing new program: /tmp/hello [Switching to process 16827] Catchpoint 1 (exec'd /tmp/hello), 0x44031840 in ?? () (gdb) --3MwIy2ne0vdjdPXF Content-Type: text/plain; charset=us-ascii Content-Disposition: attachment; filename="gdb-6.5-follow-exec.patch" Content-length: 6552 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=182116 diff -rup gdb-6.5-orig/gdb/inf-ptrace.c gdb-6.5-follow_exec/gdb/inf-ptrace.c --- gdb-6.5-orig/gdb/inf-ptrace.c 2006-01-24 23:34:34.000000000 +0100 +++ gdb-6.5-follow_exec/gdb/inf-ptrace.c 2006-07-29 11:26:32.000000000 +0200 @@ -166,8 +166,10 @@ inf_ptrace_mourn_inferior (void) /* Wait just one more time to collect the inferior's exit status. Do not check whether this succeeds though, since we may be dealing with a process that we attached to. Such a process will - only report its exit status to its original parent. */ - waitpid (ptid_get_pid (inferior_ptid), &status, 0); + only report its exit status to its original parent. + WNOHANG is required as on follow_exec() (due to TARGET_WAITKIND_EXECD) + resulted from waitpid(2) and it called us by target_mourn_inferior(). */ + waitpid (ptid_get_pid (inferior_ptid), &status, WNOHANG); unpush_target (ptrace_ops_hack); generic_mourn_inferior (); diff -rup gdb-6.5-orig/gdb/infrun.c gdb-6.5-follow_exec/gdb/infrun.c --- gdb-6.5-orig/gdb/infrun.c 2006-07-28 23:30:22.000000000 +0200 +++ gdb-6.5-follow_exec/gdb/infrun.c 2006-07-29 19:06:06.000000000 +0200 @@ -47,6 +47,7 @@ #include "language.h" #include "solib.h" #include "main.h" +#include "objfiles.h" #include "gdb_assert.h" #include "mi/mi-common.h" @@ -109,10 +110,10 @@ int sync_execution = 0; static ptid_t previous_inferior_ptid; /* This is true for configurations that may follow through execl() and - similar functions. At present this is only true for HP-UX native. */ + similar functions. */ #ifndef MAY_FOLLOW_EXEC -#define MAY_FOLLOW_EXEC (0) +#define MAY_FOLLOW_EXEC (1) #endif static int may_follow_exec = MAY_FOLLOW_EXEC; @@ -375,6 +376,37 @@ follow_inferior_reset_breakpoints (void) insert_breakpoints (); } +static void +follow_exec_restore_execfile (void *filename_new_untyped) +{ + char *filename_new = filename_new_untyped; + + /* filename_new == NULL is not expected. */ + if (filename_new == NULL && exec_bfd != NULL) + exec_file_clear (0); + /* exec_bfd == NULL is not expected. */ + if (filename_new != NULL && + (exec_bfd == NULL || strcmp (get_exec_file (0), filename_new))) + exec_file_attach (filename_new, 0); + + free (filename_new); +} + +static void +follow_exec_restore_symfile (void *filename_new_untyped) +{ + char *filename_new = filename_new_untyped; + + /* symfile_objfile == NULL is not expected. */ + if (filename_new == NULL && symfile_objfile != NULL) + symbol_file_clear (0); + if (filename_new != NULL && (symfile_objfile == NULL + || strcmp (symfile_objfile->name, filename_new))) + symbol_file_add_main (filename_new, 0); + + free (filename_new); +} + /* EXECD_PATHNAME is assumed to be non-NULL. */ static void @@ -382,6 +414,7 @@ follow_exec (int pid, char *execd_pathna { int saved_pid = pid; struct target_ops *tgt; + struct objfile *objfile, *objfile_temp; if (!may_follow_exec) return; @@ -427,6 +460,30 @@ follow_exec (int pid, char *execd_pathna error (_("Could find run target to save before following exec")); gdb_flush (gdb_stdout); + + /* During the common "run" bare command we should run again the original + * program spawning us. Stacking ordering is correct this way. */ + make_run_cleanup (follow_exec_restore_execfile, + (!exec_bfd ? NULL : xstrdup (exec_bfd->filename))); + make_run_cleanup (follow_exec_restore_symfile, (!symfile_objfile ? NULL : + xstrdup (symfile_objfile->name))); + + /* As symbol_file_add_main()->thread_db_new_objfile()->check_for_thread_db() + * would fine already loaded libpthread.so while the threading structures + * would not be yet initialized for this early inferior. + * Call before target_mourn_inferior() as it will breakpoint_re_set(). */ +#ifdef CLEAR_SOLIB + CLEAR_SOLIB (); +#else + clear_solib (); +#endif + /* Do not: symbol_file_clear()->clear_symtab_users()->breakpoint_re_set(). */ + ALL_OBJFILES_SAFE (objfile, objfile_temp) + { + free_objfile (objfile); + } + symfile_objfile = NULL; + target_mourn_inferior (); inferior_ptid = pid_to_ptid (saved_pid); /* Because mourn_inferior resets inferior_ptid. */ diff -rup gdb-6.5-orig/gdb/linux-thread-db.c gdb-6.5-follow_exec/gdb/linux-thread-db.c --- gdb-6.5-orig/gdb/linux-thread-db.c 2006-07-28 23:30:22.000000000 +0200 +++ gdb-6.5-follow_exec/gdb/linux-thread-db.c 2006-07-29 11:26:32.000000000 +0200 @@ -939,6 +939,10 @@ thread_db_wait (ptid_t ptid, struct targ || ourstatus->kind == TARGET_WAITKIND_SIGNALLED) return pid_to_ptid (-1); + /* Threading structures got reset. Return as nonthreaded. */ + if (ourstatus->kind == TARGET_WAITKIND_EXECD) + return pid_to_ptid (GET_PID (ptid)); + if (ourstatus->kind == TARGET_WAITKIND_STOPPED && (ourstatus->value.sig == TARGET_SIGNAL_TRAP || ourstatus->value.sig == TARGET_SIGNAL_ILL)) @@ -1106,6 +1110,9 @@ thread_db_mourn_inferior (void) the inferior, so that we don't try to uninsert them. */ remove_thread_event_breakpoints (); + /* Destroy thread info; it's no longer valid. */ + init_thread_list (); + /* Detach thread_db target ops. */ unpush_target (&thread_db_ops); using_thread_db = 0; diff -rup gdb-6.5-orig/gdb/testsuite/gdb.base/foll-exec.exp gdb-6.5-follow_exec/gdb/testsuite/gdb.base/foll-exec.exp --- gdb-6.5-orig/gdb/testsuite/gdb.base/foll-exec.exp 2001-03-06 09:21:50.000000000 +0100 +++ gdb-6.5-follow_exec/gdb/testsuite/gdb.base/foll-exec.exp 2006-07-29 11:26:32.000000000 +0200 @@ -45,12 +45,6 @@ if { [gdb_compile "${srcdir}/${subdir}/ } -# Until "catch exec" is implemented on other targets... -# -if ![istarget "hppa*-hp-hpux*"] then { - continue -} - proc zap_session {} { global gdb_prompt global binfile @@ -212,7 +206,9 @@ proc do_exec_tests {} { setup_xfail hppa2.0w-hp-hpux* CLLbs16760 send_gdb "continue\n" gdb_expect { - -re ".*Executing new program:.*${testfile2}.*Catchpoint .*(exec\'d .*${testfile2}).*in .START..*$gdb_prompt $"\ + # It is OS dependent and no symbols may be found, GNU/Linux has "_start" + # while HP-UX has " in .START..*$gdb_prompt" etc. + -re ".*Executing new program:.*${testfile2}.*Catchpoint .*(exec\'d .*${testfile2}).*in .*$gdb_prompt $"\ {pass "hit catch exec"} -re "$gdb_prompt $" {fail "hit catch exec"} timeout {fail "(timeout) hit catch exec"} --3MwIy2ne0vdjdPXF--